When I first got to know hadoop, I had to configure a hadoop cluster on a 7-7-8 basis. However, when I had a big hurdle, I often fell victim to the ship.
Every time you execute hadoop namenode-format to format the hadoop file system, an error is always reported. As a result, namenode cannot be started successfully after the hadoop file system is started, and it cannot be solved after you find many posts on the Internet. Later, we found that after each format operation, no matter whether it was successful or not, we deleted all the temporary hadoop files before the next format operation and then re-formatted them. It seems that the format operation was successful, I don't know.
In addition, each restart of the machine may delete all the hadoop TMP files, which may also lead to data loss in the hadoop file system. Further research is required to determine whether this is the case or why.